Add v4 image to neo4j-public/causal-cluster-k8s

[AdrienLemaire]

Not sure if this is the right place to request it.
I also wrote a more detailed request in the community forums

[AdrienLemaire]

@bfeshti I see that images 4.0 and 4.1 have now been made available on GCP kubernetes apps, thanks for that!

This time, I don't have any interesting data to restore, so I'll just drop the 3.5 cluster and recreate a 4.1 one, but what would normally be the best way to upgrade the causal cluster in that context?

[AdrienLemaire]

Actually, I cannot seem to be able to create a causal cluster 4.1 properly.

When looking at the container logs, I could find this error:

APOC couln't set a URLStreamHandlerFactory since some other tool already did this (e.g. tomcat). This means you cannot use s3:// or hdfs:// style URLs in APOC. This is caused by a limitation of the JVM which we cannot fix.

Any idea what this means and how to resolve this issue?

[bfeshti]

Hello @AdrienLemaire, thanks for letting us know about the issue with your 4.1 cluster. I have been investigating that and it was happening because Kubernetes deprecated the v1beta2 api version.
I have updated it now and it should go live in a couple of hours.
Regarding the upgrade, upgrade procedures from Neo4j documentation (https://neo4j.com/docs/operations-manual/current/upgrade/) and the README in backup and restore directories from this repository should help.
Please let me know if you have further questions.

[AdrienLemaire]

@bfeshti I tried again creating a new v4.1 cluster without luck. The same APOC error is thrown (but I suppose it's irrelevant), and there's a connection issue that ends up with a server shutdown.

My cluster node pool is auto-scalable, and I requested 3 core servers (just used the default values for everything)

[AdrienLemaire]

On this part of the log, I can see that all 3 servers got connected at 14:01:03, but 6 seconds later, the message "waiting for 3 core servers" is thrown, immediately followed by a shutdown

[bfeshti]

@AdrienLemaire The APOC warning does not affect the cluster formation.
I cannot reproduce what you are going through right now, if i go and spin up a 4.1 GKE cluster from Marketplace it works fine.
Can you check again that you dont have any disk leftovers from some old cluster ?